Long term investing (investing for one or more years) is different than trading (investing for minutes, hours, days, weeks, or a few months). The NeuroShell Trader is great for short term investing and trading, but we suggest the NeuroShell Classifier for long term stock classification. Investing for long term may involve making decisions about which of dozens or hundreds of stocks should be purchased. Furthermore, the indicators usually used to make such decisions are fundamental indicators, as opposed to the technical indicators the NeuroShell Trader is adept at manipulating.

Fundamental indicators involve the long term fundamental aspects of a company’s business, like earnings, insider stock trading, long term debt, and a number of other things. Technical indicators are more related to current prices, like open, high, low, close, volume, and functions of those over time such as moving averages.

Fundamental indicators, like dividends, often are only available quarterly, making them very difficult to use in timing models such as those in the NeuroShell Trader where the largest bar is monthly.

The NeuroShell Classifier is perfect for longer term investing because it can classify each of hundreds or thousands of stocks as a buy or no-buy. You get a probability for each stock, and you simply buy those with the highest “buy” probabilities, and sell those with the highest “no-buy” probabilities. Alternatively, you could also make a separate net for selling, where the outputs are “sell” or “hold”.

You can build your data in a spreadsheet. The columns (inputs) are fundamental indicators, except the last column (output) which is either “buy” or “no-buy”. You could even sneak some technical indicators in the input columns like price/(52 week high).

The rows (historical examples) are all the stocks in which you are interested and for which you can get fundamental data. You can include multiple rows with the same stock if each one is for a different quarter. That is, you can include any previous quarters of any stock, each quarter on a different row and each with its own output.

The output, “buy” or “no-buy”, can be determined any way you like, but the easiest and most obvious way is to set it based on whether or not the stock price went up X percent over the Y quarters FOLLOWING the date that the fundamental indicators (inputs) were measured. You can determine for yourself what X and Y should be.

Note that used this way the Classifier doesn’t plot a price time series like the NeuroShell Trader does, and doesn’t have any built-in indicators like the Trader. It also has no automatic way of downloading historical data like the Trader does, so you’ll probably get your indicators as published quarterly from research web sites or publications. You may have to do some spreadsheet work to get your historical data in order.

If you don’t have at least 200 rows we suggest that you use the genetic training method, because it has the “one hold out” feature which is so important if you don’t have many examples.